The FDA guidance on process validation is a good reference about Process Optimization Studies. In it the FDA is emphasizing data driven decisions (hence the emphasis on statistics).
I think the difference between optimization and validation, is that in optimization you don’t have acceptance criteria and pre-scribed tests cases. In validation you do have prescribed acceptance criteria. So optimization is an information gathering practice, and validation is a confirmation practice (do not explore, confirm what you already know).
Because of this optimization studies are less strict and stringent. But they should still be documented using GDP practices.
Now that being said, I think optimization should be based on knowledge already - what could impact your process (what are likely critical process parameters). Basically the optimization could follow any/all of the following steps.
1 - Brainstorm - what MIGHT affect your process
2 - Screen test - what does have some impact on your process
3 - Optimization test - what is the best operating point or range
4 - What controls do you want to have set points (do not deviate)
5 - What controls do you want to be able to adjust as needed.
6 - Once you have these things established, you can define the operating ranges/set-points, and then validate those items.